Topic: Where is the difficulty in creating an ASIC? (Read 3286 times)

There are other "ASIC" Intellectual Property out there already that can do the same computation.  However, those IPs are optimized for low area and low power, so that they can sell it to someone who wants to add hashing capability to their ASIC.


This is the key point.  IP out there are meant to do a few hashes.  So the tech sheets I've seen from IP providers calculate a hash in 66 clock cycles.  Furthermore, because they are minimizing area used, they reuse as much as possible during these 66 clock cycles, meaning you can only calculate one hash every 66 clock cycles.  Assuming you are running it at 660MHz, then you only get 5MH/s per IP core (bitcoin hash is SHA256(SHA256(x))).

But to the mobile phone ASIC designers, they are happy to license it so that they can calculate a hash in 100ns, compared to however long it would take to put it through the MIPS or ARM core, or having to implement one themselves.

In comparison, the initial open source FPGA design unrolled the SHA256 cores (if the FPGA can fit it), which means it computes SHA256 in one clock cycle and a bitcoin hash in 2 clock cycles.  It uses up a lot of area, but can roughly get MHz -> MH/s (because of pipelining, the first core can start calculating the next hash while the second one finishes)

Yep. Typical chips designed to accelerate operations like hashing are designed to accelerate the hash of a large object. An ASIC designed to accelerate Bitcoin mining gets much of its performance benefit from being designed to do precisely the hash operations that Bitcoin mining requires in precisely the sequence Bitcoin mining requires them to be done. A Bitcoin mining ASIC is an assembly line with many nonces in various states of hashing at the same time.

There has been some talk of upcoming "crypto chips", meant for use in mobile phones, being used for mining.

While their design will surely prioritize performance-per-watt highly, it will probably not prioritize absolute performance very highly at all.  (How much sha256 crunching will a mobile phone need to do, really?)

So, I think that any "miner product" that uses these chips will probably have to incorporate a heck of a lot of them in order for it to add up to any kind of substantial hashrate, which will make for a complex PCB, complex cooling, and such.

Actually I do, but it is the incorrect use of the acronym ASIC that is already being propagated that gets the terminology all backwards. People are calling the new products 'ASIC Miners' because they are mining rigs/devices that use ASICs as opposed to CPU or GPU. but the device themselves should probably have some other name given to them. Kinda like people say 'Mining Rig'... then is your 'Rig a GPU rig or an ASIC rig.

My original post was meant to say that aren't there other ASICs already designed by someone else that can do these same types of computations.

Evidently as I have learned through this thread.... no these ASICs that are being designed by BFL and AVALON are specific to BitCoin... not just AN ASIC that can do the number crunching of things(hashes) like what is used for bitcoins.

Clear as mud?

I guess OP does not even know what ASIC means

Aside from all the bridge burning Josh did, there is also the NRE to be considered for BFL, that would probably keep them from just packaging up Avalon chips into a nice little release for us. A shame, as this would really have worked out nicely for everyone.

Thread please. TY in advance.

I agree.

Avalon has already announced their intent to sell bare chips to "integrators" for exactly this kind of thing.

However I believe Yifu has said that he would first commit Seppuku before doing business with BFL, so I guess the products will have to come from other makers.

https://bitcointalksearch.org/topic/diy-pcb-with-avalon-the-quarter-stick-needs-help-161715

The biggest problem is lead time. You can't sell an ASIC for more than the value of the Bitcoins it mines. If you start developing an ASIC today, you're probably at best looking at getting to market in a year. You're also looking at having over a million dollars invested. If two other companies are already selling ASIC-based miners and the difficulty has shot through the roof, that's a million dollars you probably aren't getting back.

Suppose you start developing today and next year the difficulty has gone up by a factor of ten and the price has halved?

yup - your question just resembled what all the corporate zombie managers I see all the time where they catch wind of something 5 years late and say a bunch of rah rah spin that the job can get done better, cheaper if we just LEAN it with some 6 Sigma Agile & ITIL it NOW, cause well they want it and stuff

as everyone else said, it is hard a nut to crack and then it is a race with a short window to make $$$..  that is why Avalon is saying F it and mining themselves..

I have experience with this model in products I deal with my job. There is a manufacturer named LONmark that made a chip and a protocol on how to communicate between various electronic devices. They then license the use of the chip to manufacturers all over the world that build many different types of products that want this communication ability.

In the Mining Rigs it would be far more limiting, but could still be profitable. Especially, when the ability to MAKE money
(no pun intended) is connected with the use of a product. People can see the logic in spending more money if it will be earned back.

That is an interesting business model, kind of the ARM business model. Have a firm develop mining ASIC IP, then license it to others to manufacture. ARM chips are in your android and iphones, but but ARM holdings doesn't manufacture them, just license the tech to TI, Samsung, Apple, HTC, etc.

This works when there are firms interested in in-licensing, right now the market is a little uncertain as to how many ASIC mining companies can be supported simultaneously.

Ask BFL. They have a farkton of money and can't do it...

No problem; your question was very valid and relevant considering the extreme demand of ASIC-based bitcoin miners.

I think BFL actually outsourced their ASIC design which was probably one of the factors that led them to the trouble they're in now.  It seems like BFL figured that they'd just pay someone else to deliver ASIC chips and then only deal with building the chips into a consumer product.
On the other hand I think Avalon did both their chip design and product integration in-house, and I'm sure them being in China (geographically close to chip manufacturers) helped a lot as well.  
The BFL miners look like the iPods of bitcoin while the Avalon miner looks more like your "jumper wires and shoeboxes" project... except the Avalon actually exists and has been shipping almost according to schedule since the beginning of February, while BFL finally released a significantly-under-spec prototype 6 months after they said they'd be shipping to consumers.  
Back to the topic: Outsourcing can apparently have negative impacts. :p

**note: I didn't confirm any of the above.  It's what I'm assuming/might have read somewhere else on these forums.  So take it with a grain of salt since it may not be 100% correct.

Basically yeah; I think it can cost 10s of thousands of dollars (or more) to re-work an ASIC design once the chips have already started being fabricated.  That's why it's important to test extensively before sending a design off to be manufactured.  

Either way the difficulty comes from finding the right people to design the chips, and having the money to fabricate the chips.  You can solve both of these with money but it's a very significant amount of money, and those who can do it cheaper (i.e. Avalon, ASICMINER) will probably come out ahead.  

Thanks for the well written response... so after searching some of the terms you mention and other searches regarding manufacturing ASICs... I am better informed of what you said about the chip being the hinge point.

To me it would have been a better approach for someone like BFL and Avalon to work together on making the CHIP and then create products around that chip. I'm sure there are some differences in the firmware and hardware that would differentiate the various makers of BitCoin mining ASICs who would then pay a per unit price for use of the chip that incorporates the design costs etc.

I worry that after all these ASICs come out that they didn't get it 'right' and they don't perform like they should and have to go back to the drawing board incurring more costs the companies can't swallow. *ahem* BFL 

Where is the difficulty in creating an ASIC?

I get your smart-assness.... I'm the same way. So I'm not gonna whine. LOL

I sort of am coming from that angle... Meaning I would think there was some other chip other than those in GPU etc that would be suited for hashing.

But as someone else in this thread just explained and clarified. The chip itself is specifically designed for hashing bitcoins, therefore no, there is no other chips out there that could do specifically what a BitCoin miner is looking for.

Ranking in order of "easiness" to produce from high to low:
CPU - $0 start up
GPU - near $0 start up
FPGA - $10,000's start up capital
ASIC - $1M+ start up capital

Ranking in order of profitability:
ASIC - custom designed chip
FPGA - custom designed firmware on available chips
GPU - custom code running on widely available graphics cards
CPU - custom code running on every computer

Funny how technology works out like that... (sarcasm)

I think you are coming from a 'there's an app for that' point of view

maybe you can just outsource this?

Understood, I'm just saying that it's still a risky business move since it takes a long time to design and fab a working ASIC and BTC is very volatile.

I don't know much about hardware design, but I believe ASICs are designed using a hardware description language (HDL) such as Verilog or VHDL (this is like the "source code" of the chip).  That design is then tested and simulated and finally taken to a chip fabrication plant who actually stamps the design out into silicon chips.  The HDL code can't be reverse-engineered from an existing chip, so you need to design your own unless someone is willing to sell (or give) you a working design.  I'd assume there aren't any existing chips that will perform the same function as the Avalon, BFL, and ASICMINER chips otherwise it would be somewhat trivial for someone to buy the chips and design their own miner around them.

Yes- the ASIC chip is fundamental difference between ASIC mining hardware and every other type of mining hardware.  These are custom chips designed specifically for bitcoin mining and aren't available anywhere else.  The rest of the stuff (circuit boards, controllers, etc.) are just common parts available to everyone. (or parts that can be designed and manufactured relatively easily for a low cost, e.g. printed circuit boards)

Of course other chips can do the same operations. A 1000 times slower and less efficient.

Ok maybe you misunderstand... I am coming from a business point of view.

So these ASICs are actual custom chips that have to be created from schematics? In this digital age of computers and technology, you are saying that now other chip in the world can do similar operations as the chips Avalon and BFL have supposedly created from scratch?

It just seems there would be a much easier way to come up with a product besides creating the actual physical chip from scratch.

So the chip is the stumbling block then?

Designing an ASIC chip is very difficult simply because it requires a great deal of hardware design skill to design the chip, and a lot of money to actually produce it. (and actually a lot of money to design and simulate it as well, since the required software is extremely expensive)

Producing an ASIC chip requires a lot of money ($100k+?) because the initial setup (creating the mask, setting up all the equipment for that particular chip) is very expensive and the companies that produce the chips will only make them in huge batches due to this.

Producing an ASIC isn't a "do it yourself at home" type of project... that's what FPGAs are for.

Obviously money can solve all of these problems (you can just hire someone to do all the work if you have enough money) but it's a huge risk considering the amount of time it'd take from conception to product (just look at the existing BTC ASIC companies) and Bitcoin's extreme volatility.  There's the possibility of starting now and BTC being worthless by the time you've spent $100k on the project.

The money is obviously there. There is a market.

So why is everyone having to deal with one manufacturer who is shipping units direct from China and essentially has a monopoly on the market for now.

The other manufacturer has the smarts to know what needs to get done to make an ASIC work, but maybe not work as good as it should. Plus they must not have any logistic or business know how to be so far off on their delivery of the product as marketed and alienating their customers.

Money solves most problems... especially when they sold all these things pre-paid and got all that money upfront. I wish my industry couold get paid for goods and work BEFORE delivering it.

So where is the difficulty?

Why aren't more little startups coming up with ASICs made with jumper wires and shoeboxes? If they produce BTC... Mass produce it and SELL IT!